Outsourcing provider CRISIL Global Research and Analytics issued a report arguing that research providers will need to invest in new artificial intelligence technology in order to reduce costs and improve margins. The paper, “Investment research faces technological disruption, profitability pressures” predicts increased use of alternative data and automation of maintenance research, models and reports.
Unstructured alternative data is becoming increasing critical for differentiated research insights. At the same time natural language processing (NLP) systems have been developed to identity, extract, classify, translate, and summarize data more efficiently.
Machine learning algorithms designed to detect patterns increasingly mimic human cognitive skills, allowing automated extraction of investment ideas and actionable trading recommendations. Combined with rapid advances in computing power (gigaflops to petaflops), access to vast amounts of data (gigabytes to petabytes), advent of cloud computing, availability of open source AI packages and a significant decline in the cost of computing, the technology is now upon us.
Although investment banks have invested in automating back office operations, they have been slow to automate front office functions such as research. For more than two decades, core research functions have remained largely untouched by technology as research teams use Excel for modeling, email for report distribution, PowerPoint for presentations, and the phone for client interactions. CRISIL predicts that research solutions powered by robotics and artificial intelligence (AI) will be among the top five priorities for CIOs and research heads over the next three years.
Partly this will be a reaction to the decreasing profitability of cash equities which CRISIL says has been declining 3% annually over the last 3 years (CRISIL affiliate Coalition said bulge bracket cash equities revenues actually declined 17% in 2016). CRISIL believes that 45% of research activity is spent on maintenance research and research group profitability could be improved by 310 basis points by automating half of maintenance research activity.
The other factor impacting the adoption of new technology is research unbundling associated with MiFID II, which will place greater emphasis on differentiated research. Over the medium term, CRISIL predicts asset managers will optimize their research procurement by leveraging bulge bracket firms “for their breadth of coverage, alpha ideas, superior client servicing and advanced analytical tools” (some of CRISIL’s best clients are bulge banks); boutique/regional firms for their differentiated insights; and third-party service providers for bespoke research.
The net result will be transformation of research process and product. CRISIL foresees automation of financial models as well as structured research tasks such as maintenance research, and semi-automated reports that potentially displace junior analysts and help expand research coverage.
CRISIL believes research firms will fight for access to unstructured alternative data, which will be become a part of their competitive offering: “Sell-side providers are likely to partner with third party vendors to secure access to proprietary data. Further, firms are increasingly likely to use techniques such as web scraping and surveys and related analytics to differentiate. Social media analytics, which to date has found scant use in research floors despite considerable buy-side interest, is also likely to find a space.”
Nevertheless, CRISIL believes that research departments are resistant to change. “We observe high resistance among analysts to shift to intelligent web-based platforms that automate low-value, structured tasks.” In addition, many firms have been holding back while waiting for technologies to mature. There are also organizational challenges in creating cross-functional teams with technological and domain expertise to clean and validate data and transform it into investment intelligence.
ResearchWatch readers know that we agree that alternative data is one of the most important new trends impacting research. We also concur with CRISIL that most of the sell-side has up until now been relatively slow to embrace alternative data. Although implementation difficulties have contributed to this hesitation, a major factor in our view is economics. Bundled research pricing is a disincentive to creating new products because it is difficult to be paid incrementally for incremental new services. Only when research pricing becomes more granular, a trend many investment banks are fighting tooth and nail, will the sell side be able to fully embrace radical new products like alternative data.
So for now CRISIL’s argument for automation is predicated on the assumption that nearly half of research activity is low-value maintenance research. Automation will yield compensation savings as fewer analysts become more efficient, increasing margins. But this is an old argument, and the sell-side has already outsourced low-level maintenance research to firms like CRISIL which says it supports 75 sell-side firms with coverage of 3,300 stocks. Perhaps automation can reduce those outsourcing costs further and reduce the number of junior analysts that firms employ, but it is not clear that these savings will substantial enough to offset the technology investment required for artificial intelligence.
Moreover, alpha-rich alternative datasets by their nature do not lend themselves to sell-side distribution. Owners of alternative data can make more money skimming high rents from the buy-side rather than cheapening the value of their products with broad distribution through the sell-side. However, as alternative data products begin to become beta products — used widely enough they have little alpha but still necessary because they are used widely and impact market movements — they will lend themselves to sell side dissemination. Social media derived sentiment analysis might be one example.
Nevertheless, AI disruption of research is coming as the technology continues to mature and implementation costs decline, and the technology will be even more transformative than CRISIL suggests. It is not just junior analysts who are at risk, but senior analysts doing tried and true (and increasingly obsolete) fundamental research. Current research product will be increasingly commoditized and even the high value products – models, events, analyst access – will be at risk. As Blackrock embraces quantamental investing and the discretionary institutional management industry is being transformed by new technologies, the sell side will have no choice to adapt, or become increasingly irrelevant.